Everyday puzzles in QuIP research: Is God an explanation?
8 February 2021 | QuIP Articles|
Does every explanation of a change have to be a change itself? Can only changes explain changes?
At BSDR and Causal Map Ltd we spend a lot of time battling with tricky conceptual questions and puzzles. Working here can seem more like being part of a philosophers’ holiday camp than a research team! Sometimes I think it’s the simplicity of the QuIP approach which kicks up so many questions. Here’s one question which comes up a lot.
In QuIP, we ask people to describe changes in their lives over (say) the last three years, and then generate a backwards chain of causal explanations by asking “and what influenced that? …. and what influenced that?”. So, do all the explanations also have to be expressed as changes in the last three years? For example, someone might say this:
I know more about water conservation now because of the radio broadcasts sponsored by Organisation X, which we didn’t have three years ago. But I also know more because of the local government’s agricultural officers who explain things to me. They’ve been around forever, and they haven’t changed their activities, it’s just that I learn something new from them every time I meet them.
We recommend including the agricultural officers as a causal factor influencing the change in knowledge, even though their activity has not changed.
This means that we also sometimes code factors like “God” and “Unemployment” which are often mentioned as causal influences even though they may not themselves have changed. Of course, we hope that the interviewers have been trained to gently question whether the respondent really means to describe a causal influence and is not just producing an empty formula out of habit.
If we look at the same question from a perspective of quantitative statistics, we note that if we had data both from three years ago and from now (which we don’t), we would indeed observe a correlation between presence of radio broadcasts and knowledge of water conservation. On the other hand, we wouldn’t have any correlation between input from agricultural officers and knowledge of water conservation, because there is no variation in the officers’ input; it was the same all the time. This fact might lead us to feel that there is something illegitimate about our recommendation to include the officers’ input as a causal factor. We would perhaps like to have data from a parallel world in which there were no agricultural officers over the whole three years, to see whether the knowledge increased, but we don’t. But if we think more carefully, we will realise that nor do we in fact have data from three years ago on the radio broadcasts either. What we have in both cases is not a statistical contrast but our respondents’ more or less implicit causal claim that it was both the addition of the radio broadcasts and the presence (rather than the absence) of the agricultural officers which each made a difference.
QuIP takes the view that the validity of people’s causal information comes primarily from a whole shared knowledge map gained over time from culture, instruction and experience. For example, it is not usually the case that I think “my knowledge seems to be going up, what could have caused that?” but rather we are well aware of the causes because we are part of, inside, the whole process, and we know what it is like to gain understanding when and because someone explains something. It is true that this knowledge is sometimes updated using systematic observation of contrasting cases, whether before-against-after, or here-against-there, or even occasionally using experimental manipulation; but this is an important additional option rather than relying solely on a primary or original source of causal information.
To comment on this post please connect with us on Twitter @bathsdr
— Bath SDR (@BathSdr) February 8, 2021